SkillTester: Benchmarking Utility and Security of Agent Skills

By: Leye Wang, Zixing Wang, Anjie Xu

Published: 2026-03-28

View on arXiv →
#cs.AI

Abstract

This technical report presents SkillTester, a tool for evaluating the utility and security of agent skills. Its framework combines paired baseline and with-skill execution conditions with a security probe suite. Grounded in comparative utility and user-facing simplicity principles, it normalizes raw execution artifacts into utility and security scores and a three-level security status label. It aims to be a comparative quality-assurance harness for agent skills in an agent-first world.

FEEDBACK

Projects

No projects yet